• Nenhum resultado encontrado

Chapter 5 – Discussion, Conclusions and Future Perspectives

1. Discussion

1.2. Tackling Big Data

It has been recognized that one of the main reasons that quantification within HIA is rarely done, is that it “is often hard or impossible” because of a lack of information regarding initial conditions, effects of the proposal, and the theoretical framework linking conditions to health outcomes” 1. A major difficulty is thus to find available and reliable data to model interrelations, explicitly between health determinants, health impacts, policies and related costs 4,11,12.

The situation becomes even more complex when dealing with growing amounts of data of various sources and natures. As everyday life becomes increasingly digital and

“digitalisable” and technological progress make data storage capacity easily available, huge amounts of structured and non-structured data are being collected 13-15. The term

“Big Data” usually refers to “very large amounts of data that are routinely or

automatically collected and stored”. It is often defined by 3 characteristics known as the 3 V’s - volume, velocity and variety - or by 5 characteristics, when adding veracity and value

16.

The editor’s comment of WHO’s Eurohealth 2017 Spring issue states that “there is a growing awareness that harnessing “big data”, if done properly, could transform both the quality of healthcare for patients and how health systems perform” 17. Linking databases by sharing electronic health records and integrating other non-health care systems information sources (geographic location, socio-economic status, lifestyle and social networks) may support the transformation policy-makers, patients and providers need towards a data driven and value-based health care 13. The United Nations Global Pulse Program to “harness big data for development and humanitarian action” is also a clear statement of “how data science and analytics can contribute to sustainable development”, with projects from areas such as public health, climate and resilience, economic well-being, but also data privacy and protection or real-time evaluation18. The information used in Chapter 4 is precisely linked data at individual level from different natures and sources, namely hospital and primary care, national registries of pharmaceuticals and mortality. It consists of a pilot approach before scaling up to bigger and even more complex databases, taking advantage of the innovative health information management national approaches in Portugal 19.

Expected advantages from using this big data approach include 13:

• improving the quality of care:

o individualizing treatment plans, o decreasing duplicate diagnostic tests,

o monitoring and benchmarking provider performance;

• increasing the healthcare systems efficiency towards value-based healthcare systems:

o reducing waste from underuse of effective treatments, overuse of ineffective treatments (for example, identifying the most cost-effective treatments for each patient and hence improving patient outcomes in a cost-effective way) and failure to coordinate, manage and execute care

(for example, coordinating primary and hospital care levels towards an effective disease management),

o fast-tracking the development of innovative and more effective health technologies (for example, linked long-run real data on outcomes to assess comparative effectiveness of new medicines, leading to more informed decisions),

o creating efficiency gains in collecting and using data, analyzing available secondary data rather than primary data (decreasing information collection costs and research time).

• generating high quality research that guaranties evidence-based clinical practice and decision-making procedures (linked data at national and even international levels collected over time constitute a wealth of research possibilities, allowing researchers to work on real data about populations rather than samples, improving especially important knowledge regarding chronical non-communicable diseases).

Acknowledged difficulties to use a big data integrated approach include 13,16:

• technical challenges such as different standards used in databases that may prevent data from being comparable and compatible (for example a unique patient identifier is not always available), the complexity of linking data from different natures (many analytical tools are not appropriated to deal with this integration) and data reliability (for example missing data and error or/and bias deriving from human entry of manually fed electronic records);

• ethical challenges such as data privacy concerns (for example an opting-out option may have to be available, distinguishing consent to use data in a service provision context or in a research context) and data security issues;

• legal challenges such as diversity of legal framework between countries, even within the European Union, and between diseases (for example mandatory registries for infectious diseases, but requirements of explicit patient consent for others), that may be mitigated by assuring good practice measures are taken into account (for example establishing steering committees with patient

representatives, using trusted third parties for data linkages, creating clear rules for requesting and granting data access, as well as tracking its use);

• governance challenges demonstrated in the disparities in speed at which different countries are building big data governance frameworks to integrate technical, legal, ethical and politic aspects (even the European personal data protection regulation being updated may not keep up to the speed of continuous changes of a big data environment 20.

These difficulties are still injuring the process of further using and taking advantage of big data potential benefits 21. Nevertheless, there are international measures being taken such as the recent OECD ministerial statement to assure that countries and all stakeholders are taken on board to guaranty that big data really represents the future of health systems 22.

The HIA field has thus every chance to benefit from the increased use of big data, linked throughout health care and additional information sources related to social health determinants, obtainable for whole populations but discriminable to sub-populations and available over time.

Nevertheless, embracing big data also means an additional methodological challenge regarding analytical statistical procedures, since conventional inductive statistical methodologies are fairly limited for big data. The imminent revolution is not just the big data scaling up, but the need to develop and use new approaches to mine data from different sources and natures and extract knowledge and insights. A multi-disciplinary perspective is being adopted and the new field of Data Science is emerging and growing

23,24. Data analysis is no longer focused on outcomes related to a particular, perhaps more clinical, area of knowledge in what concerns predicting health impacts, but is present across all of science and reality.

Our current research and proposed approaches focus precisely on how to overcome big data difficulties in the point of view of statistical methodologies and thus contributes to improve quantified HIA. The statistical methodologies used throughout chapters 2, 3 and 4 (such as text mining, cluster analysis, multiple correspondence analysis, for example) are typically used in data mining and data science contexts. Moreover, they also solve

other concerns important for the HIA field: dealing with equity and generating clear graphical representations of results, as detailed in the following sections.

Documentos relacionados